Audit of outcomes in HIV

BHIVA Audit and Standards Sub-Committee

E Ong (chair), J Anderson, D Churchill, M Desai, S Edwards, S Ellis, A Freedman, P Gupta, V Harindra, A

Judd, D Ogden, O Olarinde, R Pebody, F Post, A Rodger, C Sabin, A Schwenk, A Sullivan, E Wilkins,

H Curtis (audit co-ordinator)

Aims

To assess outcomes for patients with established HIV:

In careIn and out of care.

Audit of outcomes not care qualityPatient factors such as poor adherence

explicitly not taken into account.

Aims

To test feasibility of assessing outcomes at individual sites.

Rationale: from 2013, HIV/STI audit to be procured nationally with requirements of:

Systematic annual re-audit of key outcomesPublication of site-level findings.

Methods

Case-note review in October-December 2011 of patients previously seen for HIV care during 2009.

100 patients per site to provide statistically meaningful site-level data.

Denominators based on patients:Still in care at same clinicAs above, plus not in care but not known to have died,

transferred or left UK.

Patient characteristicsN = 12975 %

Sex

Male 7794 60.1

Female 4917 37.9

Ethnicity

White 6729 51.9

Black-African 5075 39.1

Age

16-39 4979 38.6

40+ 7931 61.4

Route of HIV acquisition

Heterosexual 7404 57.1

MSM 4533 34.9

IDU 305 2.4

NB: numbers and totals do not add because of missing data and values not shown.

Case mix variation

Sites with HIV caseload >1000:Older patientsEarlier dates of ART initiation.

Small (≤100) and large (>1000) sites:More white, male, MSM than medium sitesFewer black-African, female, heterosexual.

Random and systematic case-mix variation

<100 100-1000 >10000%

100%

Site caseload

Per

cen

tag

e o

f b

lack

-Afr

ican

pat

ien

ts

Inclusion in outcomes assessment: current care status

Data submitted for 12975 patients seen for care in 2009

Exclude from analysis:1250 (9.6%) transferred care to different UK clinic

267 (2.1%) left UK

166 (1.3%) died

11292 (87.0%) “in & out of care”

Exclude from “in care” analysis:575 (4.4%) stopped attending, but not known to have transferred, left the UK or died

65 (0.5%) not known

87 (0.7%) not answered

10565 (81.4%) care at same clinic

Of which 9207 (87.1%) were on ART

10308 (79.4%) care at same clinic and seen during 2011

Main outcome measures

Planned poor outcome definitionCriteria for poor outcome Exceptions

On ART, VL >100 Recent (re)start

Probable blip, provided VL <200, or VL <1000 and measured within last 60 days

Considered not of clinical concern, provided VL <200

On ART, VL not measured within 230 days preceding audit

Not on ART, CD4 <350

Previously but not currently on ART Stopped after MTCT prevention

Not seen in clinic during 2011 Seen in 2010 and considered stable and not needing to be seen frequently

Poor outcome rates: corrected data

Poor outcome rate

Poor outcomes

Total

Patients in and out of care

12.1% 1364 11,292

Patients in care 7.1% 751 10,565

Patients in care and seen during 2011

6.4% 659 10,308

Variation in poor outcome rates by patient and site characteristics

  In and out of care In careIn care and seen

during 2011

Overall 12.1% 7.1% 6.4%

Male  10.9% 6.7% 6.0%

Female  **13.9% *7.8% 7.0%

White  10.0% 6.9% 6.3%

Black-African  **14.7% 7.2% 6.4%

MSM  9.6% 6.3% 5.6%

Heterosexual **12.9% 7.3% 6.6%

ART initiated 2005 or earlier 9.3% 6.6% 6.0%

ART initiated 2006 or later 9.6% 6.2% 5.7%

<100 site caseload <1000  12.3% 7.1% 6.5%

Site caseload >1000  12.6% 7.2% 5.8%

Site caseload <100  11.8% 7.9% 7.2%

** Significant at 99%. * Significant at 95%.

Possible outlier sites

0 2 4 6 8 10 12–100%

– 50%

0%

+ 50%

+100%

+150%

+200%

X

Y

Z

Funnel plot: corrected data, patients in and out of care

Expected count: Patients with poor outcomes

Ob

serv

ed p

oo

r o

utc

om

es r

elat

ive

to e

xpec

ted

Source: BHIVA audit 2011-12

Problems with outcome definition

Burdensome for sites to provide sufficient data.

Data quality issues – missing data and errors, especially dates and reasons for exceptions.

Action taken in response:Corrected missing data by imputing. E.g. if

attendance date given but VL date missing, assume the same

Tested simpler definition of poor outcome for potential future use

• On ART with VL >200• Not on ART, CD4 <350.

Comparison of outcome measures: poor outcome rates for patients seen for care during 2011 by site caseload

▪Simple VL/CD4 only outcome ▪planned outcome, uncorrected data ▪planned outcome, corrected data

Secondary outcomes

London CQIN viral load outcome

London≥90% adults starting ART to have <50 copies/ml within

3-15 months, excluding pregnant women and those with no available measurement (achieved 2010: 93%).

BHIVA auditNo direct comparison9070 patients in care on ART with available result,

83.2% had undetectable VL and a further 10.4% had detectable VL <100 copies/ml.

London CQIN CD4 targets

London≥95% of adults in care for at least one year at

the same site to have ≥200 cells/mm3 (achieved 2010: 96%)

≥90% to have ≥350 cells/mm3 (achieved 2010: 83%).

BHIVA AuditOf 10565 patients in care, 94.9% had >200

and 80.8% >350 cells/mm3

Percentage of patients in care & seen during 2011 with CD4 ≤200 cells/mm3 by site caseload and whether on ART

Other secondary outcomes

Cardiovascular risk monitoring

Assessment of adherence

Percentage of patients on ART, in care and seen during 2011 for whom lipids not measured in the past 3 years

Percentage of patients in care and seen during 2011 for whom blood pressure not recorded in the past 15 months

Percentage of patients on ART, in care and seen during 2011 for whom information on adherence not recorded in the past 12 months

Conclusions: main outcome

Overall HIV treatment outcomes were excellent 87.9% of patients overall 93.6% of those attending for care.

Outcomes varied widely between sites. Poor outcomes might be attributed to patient factors Not a reflection of care quality.

Reassuring lack of systematic variation in outcomes for patients attending for care

Variation among patients out of care requires further investigation Some possible outlier sites identified May partly reflect poor data quality.

Conclusions: monitoring

Recording of cardiovascular risk monitoring and adherence assessment was variable.

Conclusions: feasibility

The audit method posed problems and is not feasible for year-on-year repetition:

Participants found it burdensome to provide dataErrors and missing data affected site-level results

significantly.

Data was imputed where feasible in this audit. In the future joint HIV/STI audit programme, sites may be judged on data quality as well as underlying outcomes.

Using a simpler outcome definition gave significantly different results.

Recommendations

Preparation for the HIV/STI National Clinical Audit and Patient Outcomes Programme should focus on:

Improving outcome measures to assess quality of care with the minimum burden for participants

Addressing data quality with automated data collection being a high priority.

Acknowledgements

All the centres who participated in this audit.